Emmanuel A. Otchere, VP & Chief Technical Expert, Huawei Technologies Co. Ltd contributed this opinion piece.
What do we really mean with autonomous systems, autonomous networks and autonomous operations? How do we make sense of them? These are the ten perspectives influencing solutions, services, and standards beyond 2023.
- What is an Autonomous System (AS) vs Autonomous Network (AN) vs Autonomous Operations (AO)?
The three concepts involve the use of technology to enable independent operations, they differ in terms of the scope and complexity of operations they can enable, as well as the level of autonomy involved. Here is my simple recap of the three concepts.
- Autonomous Systems: Autonomous systems refer to “individualized” machines or devices that are capable of operating independently and making decisions without human intervention. Autonomous systems may include a wide range of technologies, such as robots, drones, or self-driving vehicles.
- Autonomous Networks: Autonomous networks refer to “networks of interconnected devices or systems” that are able to operate and coordinate their actions without human intervention. Autonomous networks may involve the integration of multiple autonomous systems and may be used to enable complex tasks or operations.
- Autonomous Operations: Autonomous operation is a reference for processes and functions that are carried out by autonomous systems and/or networks. Autonomous operations use autonomous systems and/or autonomous networks to perform operations with no human intervention. Autonomous operations will coordinate multiple autonomous systems or devices or networks to improve efficiency, reduce costs, or increase agility.
2. How will you compare and contrast an Autonomous System with an Autonomous Network?
The two, autonomous systems and autonomous networks are core to realizing autonomous operations. While the two are also related, they are not the same thing.
- A key difference between an Autonomous System and Autonomous Network is the level of complexity involved. Autonomous systems are generally limited in scope, as they are focused on the operation of a “single” device or system (please note here single in quotes). Autonomous networks, on the other hand, involve the coordination and integration of“multiple devices or systems”, which can be more complex and challenging to manage.
- Another key difference is the level of autonomy involved. Autonomous systems generally have higher levels of autonomy, as they are able to operate and make decisions completely independently. Autonomous networks, on the other hand, may rely on human intervention or oversight to some degree, as they involve the coordination of “multiple” systems or devices and may span a wide geographic footprint or locale.
3. Can Autonomous operations improve the deployment and use of autonomous networks?
Autonomous operations can be used to improve the deployment and use of autonomous networks in several ways including:
- Automating deployment and configuration: Through autonomous operations, the deployment and configuration of autonomous networks can be automated, which can reduce the time and effort required to set up and deploy these networks. This can also accelerate the deployment time and improve their operational efficiency.
- Enhancing performance and reliability: Autonomous operations can be used to optimize the performance and reliability of autonomous networks by enabling automated adaptation to changes in the environment or network conditions. For example, where an autonomous network is able to detect and respond to problems or malfunctions in real time or adjust operations to optimize performance.
- Improving scalability: Autonomous operations can be used to improve the scalability of autonomous networks by enabling them to automatically adjust to changes in demand or workload. This can help to ensure that autonomous networks are able to meet the needs of users and maintain high levels of performance even as demand increases.
4. Can you give an example of the implementation of an autonomous system, autonomous network, and autonomous operations?
Three examples of the implementation of autonomous systems, autonomous networks, and autonomous operations include:
- Autonomous systems: An example of the implementation of an autonomous system is the self-driving vehicle. Self-driving vehicles are equipped with sensors, cameras, and other technologies that allow them to navigate and operate independently without the need for human intervention.
- Autonomous networks: An example of the implementation of autonomous networks is in Smart home automation where a network of devices that are connected to each other and able to communicate and cooperate to control and automate various functions within a home. For example, a smart home system might be able to control the lighting, temperate, and security of a home, and respond to voice commands or other stimuli. Software-defined networking (SDN) and AI enable coordination of the independent parts of the “networks” to be flexible and adaptable, and also improve their performance and reliability needed as a whole.
- Autonomous operations: An implementation of autonomous operations is the use of autonomous systems and/or networks to support realizing an outcome. In the case of autonomous drones for delivery services, autonomous drones are assigned objectives to navigate and operate independently and can be used in service delivery of packages or other items to locations that would be difficult or impossible for humans to reach.
5. When applying autonomy of technology to realize autonomous operations, what are some of the key KPIs (key performance indicators) for measuring and managing such autonomy of technology (AoT)?
- Failure rate: This KPI measures the percentage of times that an autonomous technology fails to perform its intended function. The formula is: Failure rate = (Number of failures / Total number of operations) x 100
- Accuracy: This KPI measures the degree to which the output of an autonomous technology is correct or incorrect. The formula is: Accuracy = (Number of correct outputs / Total number of outputs) x 100
- Compliance: This KPI measures the degree to which an autonomous technology adheres to relevant laws, regulations, and ethical guidelines. The formula is: Compliance = (Number of compliant actions / Total number of actions) x 100
- Security: This KPI measures the extent to which an autonomous technology is protected against cyber-attacks and other security threats. The formula is: Security = (Number of successful attacks / Total number of attempts) x 100
- Impact on stakeholders: This KPI measures the effect that autonomous technology has on different stakeholders, including employees, customers, and shareholders. The formula is: Impact on stakeholders = (Number of positive impacts / Total number of impacts) x 100
- Return on investment (ROI): This KPI measures the profitability of an autonomous technology investment. The formula is: ROI = (Net profit / Cost of investment) x 100
- Time saved: This KPI measures the time saved by using autonomous technology to perform tasks or processes. The formula is: Time saved = (Time required to perform the task manually - Time required to perform the task with the autonomous system/network)
- Cost savings: This KPI measures the cost savings achieved by using autonomous technology to perform tasks or processes. The formula is: Cost savings = (Cost to perform the task manually - Cost to perform the task with autonomous technology.
6. How will you characterize the functional parts of a blueprint for the autonomy of technology (AoT)?
Characterizing a realistic blueprint depends on the applied intent and supported requirement. In a general, an overview of the components from a technology architecture viewpoint includes:
- Sensors used to collect data about the environment or system/network.
- Actuators that are used to perform actions based on the data collected by the sensors.
- Computing platform that provides the computational platform for the data collected by the sensors and controls the actions of the actuators.
- Communication network that connects the different components of the autonomous system/network, and facilitates communication and data transfer.
- Data storage and management that enables storing and managing data collected by the sensors and generated by the autonomous system/network.
- Control algorithms that are used to process the data collected by the sensors and control the actions of the actuators.
- User interface that may be used to interact with the autonomous system/network (as the case may apply)
- Safety and security protocols that are responsible for ensuring the safety and security of the autonomous system/network and the data processed. Integration with other systems: In many cases, an autonomous system will need to be integrated with other systems, such as enterprise systems or external data sources.
- Integration component of the architecture that facilitates communication and data exchange between the autonomous system/network and other systems/networks.
- Maintenance and support tools and processes that are used to maintain and support the autonomous system.
7. In business language, what will be the key resources to specify the management information of closed-loop automation?
In a closed-loop managed entity, the entity receives feedback on its actions and uses this feedback to adjust its behavior, and improve its decision-making. The business information model includes:
- Input resource: This is the “data” that is used to control the entity and make decisions. This may include data from sensors, user inputs, or other sources.
- Output resource: This is the “data”, that is produced by the system as a result of its actions. This may include data on the system's performance, such as error rates or efficiency.
- Feedback resource: This is the “data”, used to adjust own behavior based on past and present performance. It may include data on the entity's output, as well as data on external factors that may affect performance.
- Control resources: These are the “algorithms” that are used to process the input and feedback “resources” and adjust the own behavior. These algorithms may use statistical and/or machine learning techniques to continuously improve performance over time.
8. Why is closed-loop automation critical to the “autonomy of technology”?
Closed-loop automation is critical to the success of autonomy of technology operations because it enables autonomous systems/networks to continuously monitor, control, and adjust behavior based on real-time data and feedback. This allows autonomous systems/networks to adapt to changing conditions, optimize their performance, and ensure that they are meeting their set goals and objectives.
Closed-loop automation improves autonomous operations by bringing:
- Adaptability: Closed-loop automation enables the autonomous system/network to continuously monitor and adapt to changing conditions in the environment, thus being able to respond to new challenges and opportunities, and optimize performance based on real-time data.
- Efficiency: Closed-loop automation drives autonomous systems/networks to self-optimize performance and minimizes waste. For example, by continuously monitoring and adjusting behavior based on real-time data, the autonomous system/network can minimize energy consumption, reduce errors, and improve overall efficiency.
- Reliability: Closed-loop automation enables the autonomous system/network to continuously monitor and adjust behavior to ensure that it is meeting its goals and objectives with consistency and without failure. This helps to drive the reliability of the autonomous system/network and reduce the risks of failure.
- Safety: Closed-loop automation enables the autonomous system/network to continuously monitor and adjust its behavior to ensure safety. For example, by continuously monitoring and adjusting speed, trajectory, and other parameters based on real-time data, an autonomous vehicle can ensure that it is operating safely and avoiding accidents.
9. How will you describe and differentiate human-in-the-loop and human-on-the-loop concepts?
Human-in-the-loop (HITL) and human-on-the-loop (HOTL) refer to two different approaches to the integration of human decision-making in automated systems.
In a human-in-the-loop (HITL) system, humans are actively involved in the decision-making process and are able to intervene and make decisions in real time as needed. This approach is often used in systems where there is a high level of uncertainty or complexity, and where the system needs to be able to adapt and respond to changing conditions in real time. Human-in-the-loop (HITL) examples include:
- An air traffic control system, in which human controllers are responsible for making real-time decisions about the movement and routing of aircraft based on input from automated systems.
- A self-driving car, in which a human driver is able to intervene and take control of the car if necessary.
- A medical diagnosis system, in which a human doctor is able to review and modify the diagnosis made by an automated system based on additional input and knowledge.
In a human-on-the-loop (HOTL) system, humans are not actively involved in the decision-making process but are able to review and override decisions made by the system as needed. This approach is often used in systems where there is a lower level of uncertainty or complexity, and where the system is able to make decisions based on predetermined criteria or rules.
Overall, the main difference between human-in-the-loop and human-on-the-loop systems is the degree to which humans are involved in the decision-making process. In HITL systems, humans are actively involved in making decisions in real-time, while in HOTL systems, humans are able to review and override decisions made by the system, but are not actively involved in the decision-making process. Some human-on-the-loop (HOTL) examples are:
- An automated trading system, in which human traders are able to review and override trades made by the system based on predetermined criteria.
- A machine learning system, in which human experts are able to review and modify the decisions made by the system based on additional input and knowledge.
- An autonomous drone, in which a human operator is able to review and override the decisions made by the drone based on predetermined criteria.
10. What can I do to transform an Autonomous network (AN) into an Autonomous System (AS)?
An autonomous system is a single, self-contained system that is able to operate independently. It is understandable that solution engineers will like to see a network of multiple devices/systems operate as one. This is perhaps a means to ensure better mandates of governance and goals where multi-vendor systems, different standards, and different levels of automation exist in a network… amongst other concerns.
I think that transforming AN into AS would be more than transforming the network to become a single, self-contained unit capable to be operable, altogether, independently. This challenge will require a high and deep degree of integration of the various aspects of the network, perhaps an ambition that requires a high degree of maturity in standards across the realms of communication technology (CT), information technology (IT), and operations technology (OT), as well as robust integration prowess.
A typical communications service provider is applying CT, IT, and OT capabilities across its network to realize full operations. Therefore, the transformation of the communication service provider’s network – AN to be like an AS will require a high level of integration cohesiveness to achieve this.
It’s a transformation journey with a number of different approaches, assuming that it is clear what the specific characteristics and requirements for such a network are going to be. I envisage that journey to include:
- Massive need to consolidate the various components and devices that make up the network into a single, cohesive, and working system;
- Re-engineering network systems and designing them to make them more self-contained and autonomous; as well as
- Implementing automation and self-adaptation technologies that will enable the different systems to operate and adapt both independently, and collectively together for setting goals.